Exponential Hybrid Gannet Bald Leader Optimization based Deep Convolutional LeNet for Human Behavior Analysis with Dermatoglyphics Fingerprint Pattern Classification
Atul Bhimrao Mokal, Brijendra Parasnath Gupta · 2025
Naturally, humans validate many things through touch and sensory perception, often without difficulty or even conscious awareness. Analyzing human behavior is a crucial area of computer vision, attracting considerable attention owing to its applications in human-computer interaction and assisted living. Moreover, classical modules neglected to analyze the human behavior using fingerprints and effective fingerprint classification approach is significant to identify the abnormal human behavior. Thus, this research presents an Exponential Hybrid Gannet Bald Leader Optimization based Deep Convolutional LeNet (EHGBLO_DCLeNet) for Human behavior analysis with Dermatoglyphics fingerprint pattern classification. Herein, the fingerprint image is pre-processed by employing Non-Local Means (NLM) filter. Thereafter, minutiae feature, and texture-based features are extracted from preprocessed image. Then, fingerprint pattern classification is conducted using DCLeNet, and it is developed by incorporating Deep Convolutional Neural Network (DCNN) and LeNet. Here, hybridization is done using majority voting and is tuned with EHGBLO. Finally, each classified pattern is utilized for fingerprint matching to recognize human behavior. Moreover, evaluation results demonstrate that proposed model attained an improved level of performance, with 95.205% accuracy, 93.585% True Positive Rate (TPR), 95.253% True Negative Rate (TNR), 93.6% Positive Predictive Value (PPV), 92.596% Negative Predictive Value (NPV), and 6.415% False Negative Rate (FNR).